<p>To tackle the challenge of heavy-metal determination is with overlapping fluorescence responses, a multichannel fluorescent sensor array was constructed using three silver nanoclusters (AgNCs) with different surface ligands [histidine (His), poly(methyl vinyl ether-alt-maleic acid) (PMVEM), and poly(methacrylic acid) (PMAA)]. The fluorescence responses toward seven typical metal ions (Al³⁺, Cu²⁺, Fe³⁺, Hg²⁺, Cr³⁺, Cd²⁺, and Pb²⁺) at multiple concentration levels were converted into normalized spectral fingerprints and analyzed using an automated machine learning framework (AutoGluon), which integrates Top-K feature selection, multiclass ion classification, and ion-specific concentration regression. The conventional single-probe LODs ranged from 0.16 to 28.29 µM, and the AutoGluon-assisted workflow achieved 100% classification accuracy for single-ion samples while enabling ion-specific concentration prediction from the same multichannel spectral fingerprints. The array not only discriminated samples with predefined mixture ratios, but also worked well for spiked lake water and soil extracts, enabling reliable classification and low-level quantitative detection of metal ions in real environmental matrices. These results show that combining cross-reactive AgNCs with automated machine learning provides a practical framework for rapid qualitative identification and quantitative prediction of multiple heavy metal ions in aqueous and environmental samples.</p> Graphical Abstract <p></p>